memhq 0.1.0

Official Python SDK for MemHQ — a Mem0-style drop-in memory layer for AI agents.

memhq — Python SDK for MemHQ

The official Python client for MemHQ, a drop-in memory layer for AI agents. Same shape as Mem0, with a built-in synthesis pass (ask()) for cited answers.

Install

pip install memhq

Quickstart

import os
from memhq import MemoryClient

client = MemoryClient(api_key=os.environ["MEMHQ_API_KEY"])

# 1. Add — auto-creates the user + default thread
client.add(
    messages=[{"role": "user", "content": "I'm a vegetarian, allergic to nuts."}],
    user_id="user_123",
)

# 2. Search — hybrid retrieval (BM25 + vector + graph)
results = client.search("dietary restrictions", user_id="user_123")
for memory in results:
    print(memory.score, memory.content)

# 3. Ask — synthesized answer with citations
answer = client.ask("What should I avoid eating?", user_id="user_123")
print(answer.text)
for cit in answer.citations:
    print(" •", cit.content)

Async

import asyncio
from memhq import AsyncMemoryClient

async def main():
    async with AsyncMemoryClient() as client:  # picks up MEMHQ_API_KEY
        await client.add(
            messages=[{"role": "user", "content": "I live in Brooklyn"}],
            user_id="user_123",
        )
        result = await client.ask("Where does the user live?", user_id="user_123")
        print(result.text)

asyncio.run(main())

Configuration

Env var	Default	Notes
MEMHQ_API_KEY	(required)	Get one from the dashboard.
MEMHQ_BASE_URL	https://api.memhq.ai	Override for self-host or local dev.

# Self-hosted MemoryOS or local development
client = MemoryClient(api_key="...", base_url="http://localhost:3000")

Reference

client.add(messages, *, user_id, group_id=None, metadata=None) -> AddResult

Ingest one or more messages into the user's memory graph. Extraction runs asynchronously on the server — typically completes in under three seconds.

client.search(query, *, user_id=None, group_ids=None, limit=10) -> SearchResult

Hybrid search across the user's graph plus any shared group graphs. Returns an iterable SearchResult; each element is a Memory(id, content, type, score).

client.ask(question, *, user_id=None, group_ids=None, limit=8) -> AskResult

Retrieve, rerank, and synthesize a cited answer. Returns an AskResult with answer (alias: text) and a list of Citation(id, content).

client.users.get(user_id), client.users.delete(user_id), client.users.list()

User management. user_id accepts the external id you passed to add().

License

Apache-2.0.